balitsky–kovchegov evolution equation · introduction • when the center of mass energy in a...
TRANSCRIPT
Balitsky–Kovchegov evolution equation
Emmanuel Gr ave de [email protected]
High Energy Phenomenology Group
Instituto de Fısica
Universidade Federal do Rio Grande do Sul
Porto Alegre, Brazil
GFPAE – UFRGShttp://www.if.ufrgs.br/gfpae
BK, E.G. de Oliveira–GFPAE – p. 1
Outline
• Introduction• Dipole picture• Balitsky–Kovchegov equation• Solutions to BK equation• Saturation scale• GBW model• BK equation in momentum space• Traveling waves• NLO BK• Pomeron loops• Fluctuations
BK, E.G. de Oliveira–GFPAE – p. 2
Introduction
• When the center of mass energy in a collision is much bigger than the fixed hardscale (for example, in DIS, the photon virtuality is the hard scale and s≫ Q2),parton densities increase with growing energy.
• This produces larger scattering amplitudes.
• In this regime of saturation, BFKL equation is not valid anymore.
• Using the qq-dipole model, Kovchegov was able to derive a new equation.
• Kovchegov equation is not exact, but it is a mean field approximation of a hierarchyof equations derived previously by Balitsky.
• This equation is nonlinear and can be reduced to BFKL equation when thenonlinear term is disregarded.
BK, E.G. de Oliveira–GFPAE – p. 3
Dipole picture
• For simplicity, we should consider deep inelastic scattering of a virtual photon on ahadron or nucleus.
• In the dipole picture, the system is studied in the reference frame of the target (thehadron or the nucleus is at rest).
• Therefore, all QCD evolution is included in the wave function of the incomingphoton.
• The incoming photon cannot directly strongly interact with the target.
• The splitting of the photon into a quark-antiquark pair is then considered.
• This heavy quark-antiquark pair is called onium.
• Quarks have color, but the incident photon is colorless.
• Then, the quark has color and the antiquark the correspondent anticolor.
BK, E.G. de Oliveira–GFPAE – p. 4
Dipole picture
• In the figure, the photon splitting is shown.
• z1 is the fraction of light cone momentum.
• x0 and x1 are quark and antiquark positions (x01 = x0 − x1).
BK, E.G. de Oliveira–GFPAE – p. 5
Gluon emission
• The quark or antiquark can emit soft gluons (z2 << z1).
• These soft gluons can also emit softer gluons.
BK, E.G. de Oliveira–GFPAE – p. 6
Planar diagrams
• Each loop count as Nc.
• Each vertex count as αs.
• Therefore, planar diagrams are enhanced by N2c compared to nonplanar diagrams
of the same order em αs.
BK, E.G. de Oliveira–GFPAE – p. 7
Large-Nc limit
• In the limit of large number of colors (Nc → ∞), these soft gluons can beconsidered as quark-antiquark pairs.
• Therefore, the original dipole (x01) splits into two new dipoles (x02 and x21).
BK, E.G. de Oliveira–GFPAE – p. 8
The onium
• The original incoming photon can be represented now by an arbitrary number ofdipoles.
• These dipoles do not interact with each other.
BK, E.G. de Oliveira–GFPAE – p. 9
Onium-hadron scattering
• Now, the nucleus or hadron is not going to scatter off the photon, but off thecomplex structure of color dipoles:
pp
• Two gluons must be exchanged to conserve color.
BK, E.G. de Oliveira–GFPAE – p. 10
Single and multiple scattering
• Single dipole scattering leads to linear BFKL equation.
pp
• Multiple dipole scatterings lead to nonlinear BK equation.
pp
BK, E.G. de Oliveira–GFPAE – p. 11
Fan diagrams
• Pomeron merging in the target can be understood as a multiple scattering betweentarget and projectile.
≡
• Therefore, all fan diagrams are included.γ
N
*
BK, E.G. de Oliveira–GFPAE – p. 12
Diagrams not included
• Not all diagrams are included, however.
• Diagrams representing Pomeron merging (a) cannot be represented as a multiplescattering (b).
AAA
(a)
* *
(b)
γγ
BK, E.G. de Oliveira–GFPAE – p. 13
F2 structure function
• In the dipole picture, the F2 structure function is given by:
F2(x,Q2) =Q2
4π2αem
Z
d2x01dz
2πΦ(x01, z)σdip(x01, Y ) (1)
where Φ(x01, z) = ΦT (x01, z) + ΦL(x01, z) is the sum of the square of thetransverse (T) and longitudinal (L) photon wavefunctions (more specifically, thewavefunction for a photon to go into a dipole)
ΦT (x01, z) =2NcαEM
πa2K2
1 (x01a)[z2 + (1 − z)2] (2)
ΦL(x01, z) = 4Q2z2(1 − z)2K20 (x01a). (3)
with a2 = Q2z(1 − z).
• σdip is the dipole-nucleus (or nucleon) cross section
BK, E.G. de Oliveira–GFPAE – p. 14
The forward scattering amplitude
• The dipole cross section is given by
σdip(x01, Y ) = 2
Z
d2b01N (b01,x01, Y ). (4)
• The quantity b01 is the impact parameter given by:
b01 =x1 + x0
2(5)
• N (b01,x01, Y ) is the propagator of the quark-antiquark pair through the nucleusrelated to the forward scattering amplitude of the quark-antiquark pair with thenucleus.
BK, E.G. de Oliveira–GFPAE – p. 15
Balitsky–Kovchegov equation
• The Balitsky–Kovchegov equation is given by:
dN (b01,x01, Y )
dY=
αs
2π
Z
d2x2x2
01
x220x
212
h
N“
b01 +x21
2,x20, Y
”
+N“
b01 +x20
2,x21, Y
”
−N (b01,x01, Y )
−N“
b01 +x12
2,x20, Y
”
N“
b01 +x20
2,x21, Y
”i
(6)
• This equation evolves N (b01,x01, Y ).
• The evolution quantity is the rapidity Y ≈ ln1/x.
• Besides the evolution variable, one has 4 degrees of freedom (2 from b01 and 2from x01).
• It resumms all powers of αsln1/x.
• αs = αsNc/π
BK, E.G. de Oliveira–GFPAE – p. 16
Balitsky approach
• A general feature of high-energy scattering is that a fast particle moves along itsstraight-line classical trajectory and the only quantum effect is the eikonal phasefactor acquired along this propagation path.
• In QCD, for the fast quark or gluon scattering off some target, this eikonal phasefactor is a Wilson line:
• the infinite gauge link ordered along the straight line collinear to particle’s velocitynµ:
Uη(x⊥) = Pexpn
ig
Z ∞
−∞du nµ A
µ(un+ x⊥)o
, (7)
• Here Aµ is the gluon field of the target, x⊥ is the transverse position of the particlewhich remains unchanged throughout the collision, and the index η labels therapidity of the particle.
• Repeating the argument for the target (moving fast in the spectator’s frame) wesee that particles with very different rapidities perceive each other as Wilson lines.
• Therefore, Wilson-line operators form the convenient effective degrees of freedomin high-energy QCD.
BK, E.G. de Oliveira–GFPAE – p. 17
Balitsky approach
• At small xB = Q2/(2p · q), the virtual photon decomposes into a pair of fastquarks moving along straight lines separated by some transverse distance.
• The propagation of this quark-antiquark pair reduces to the “propagator of thecolor dipole” U(x⊥)U†(y⊥) - two Wilson lines ordered along the direction collinearto quarks’ velocity.
• The structure function of a hadron is proportional to a matrix element of this colordipole operator
Uη(x⊥, y⊥) = 1 − 1
NcTr{Uη(x⊥)U†η(y⊥)} (8)
• sandwiched by target’s states.
• The gluon parton density is approximately
xBG(xB , µ2 = Q2) ≃ 〈p| Uη(x⊥, 0)|p〉
˛
˛
˛
x2⊥
=Q−2(9)
where η = ln 1xB
.
BK, E.G. de Oliveira–GFPAE – p. 18
Balitsky approach
• The energy dependence of the structure function is translated then into thedependence of the color dipole on the slope of the Wilson lines determined by therapidity η.
• At relatively high energies and for sufficiently small dipoles we can use the leadinglogarithmic approximation (LLA) where αs ≪ 1, αs lnxB ∼ 1 and get thenon-linear BK evolution equation for the color dipoles:
d
dηU(x, y) =
αsNc
2π2
Z
d2z(x− y)2
(x− z)2(z − y)2[U(x, z)+ U(y, z)−U(x, y)−U(x, z)U(z, y)]
(10)
BK, E.G. de Oliveira–GFPAE – p. 19
Balitsky hierarchy
• The operator equation creates an hierarchy of equations for the observables:
d
dη
D
U(x, y)E
=αsNc
2π2
Z
d2z(x− y)2
(x− z)2(z − y)2
hD
U(x, z)E
+D
U(y, z)E
−D
U(x, y)E
−D
U(x, z)U(z, y)Ei
(11)
d
dη〈Txy〉 =
αs
2π
Z
d2z M(x, y, z) [〈Txz〉 + 〈Tyz〉 − 〈Txy〉 − 〈TxzTzy〉] (12)
• In the mean field approximation, 〈TxzTzy〉 = 〈Txz〉〈Tzy〉, and BK equation isrecovered.
BK, E.G. de Oliveira–GFPAE – p. 20
The BFKL limit
• The last (and nonlinear) term is a direct consequence of taking into accountmultiple scatterings.
• BK equation is a closed equation, instead of Balitsky original hierarchy ofequations.
• The mean field approximation of Balitsky first equation gives the BK equation.
• If we disregard the last term, we obtain the following linear equation that can beshown equivalent to BFKL equation:
dN (b01,x01, Y )
dY=
αs
2π
Z
d2x2x2
01
x220x
212
h
N“
b01 +x21
2,x20, Y
”
+N“
b01 +x21
2,x20, Y
”
−N (b01,x01, Y )i
(13)
BK, E.G. de Oliveira–GFPAE – p. 21
BK versus AGL equation
• BK, AGL and GLR equations sum the same diagrams (for example):
• All three use the eikonal approximation, i. e., incoming photon momentum haslarge q+.
• However, BK equation does not consider large Q2. It resums all terms1/(2q+q−) ≈ 1/Q2.
BK, E.G. de Oliveira–GFPAE – p. 22
BK in 0+1 dimensions
• If one neglects the spatial dependence (N (x02,b0, Y )), BK equation is given bythe logistic equation:
dNdY
= ω(N −N 2) ω > 0. (14)
• This equation has two fixed points, one unstable (N = 0) and another stable(N = 1).
• For every initial condition in 0 < N ≤ 1, for Y → ∞, N → 1.
• In other words, there is saturation for highY (small x).
• If the N 2 term is neglected, the fixed point at N = 1 disappears and N → ∞ forY → ∞.
BK, E.G. de Oliveira–GFPAE – p. 23
BK in 0+1 dimensions
• The logistic equation complete solution is (C = N (Y = 0)):
N (Y ) =CeωY
1 + C(eωY − 1). (15)
• Solutions: linear (red) and nonlinear (blue) equations.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0 1 2 3 4 5 6 7 8 9 10
C=0.01
Y
N(Y
)
BK, E.G. de Oliveira–GFPAE – p. 24
BK in 1+1 dimensions
• If impact parameter is disregarded and only sizes of dipoles are considered:
N (b,x, Y ) → N (r, Y ). (16)
• BK equation is then simplified:
dN (|x01|, Y )
dY=
αs
2π
Z
d2x2x2
01
x220x
212
[N (|x20|, Y ) + N (|x21|, Y )
−N (|x01|, Y ) −N (|x21|, Y )N (|x20|, Y )] (17)
• Physically speaking, this represents the scattering on infinite and uniform nucleus.
• We note that the kernel already depends only on dipole sizes.
BK, E.G. de Oliveira–GFPAE – p. 25
BK in 1+1 dimensions
• The same kind of solutions of BK equation in (0+1) dimensions is found for BKequation in (1+1) dimensions for fixed r.
• Solutions: linear (red) and nonlinear (blue) equations.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2 4 6 8 10 12 14 16 18
r=0.48
r=4.8
Y
N(Y
,r)
BK, E.G. de Oliveira–GFPAE – p. 26
BK in 1+1 dimensions
• Solutions for different initial conditions in linear (above) and log (below) scales.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 2 3 4 5 6 7 8 9 10r=|x01|
N(Y
,r)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 2 3 4 5 6 7 8 9 10r=|x01|
N(Y
,r)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 2 3 4 5 6 7 8 9 10r=|x01|
N(Y
,r)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
10-2
10-1
1 10r=|x01|
N(Y
,r)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
10-2
10-1
1 10r=|x01|
N(Y
,r)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
10-2
10-1
1 10r=|x01|
N(Y
,r)
BK, E.G. de Oliveira–GFPAE – p. 27
Saturation scale
• The figures shown previously can be divided into three regions in r:
◦ The small r region, where nonlinear corrections are negligible;
◦ The large r region, where nonlinear corrections dominate and N ≈ 1;
◦ The region between the first two.
• One can introduce the saturation scale:
r <1
Qs(Y )→ N << 1,
r >1
Qs(Y )→ N ≈ 1.
BK, E.G. de Oliveira–GFPAE – p. 28
Saturation scale
• One can get a saturation scale from BK solutions.
• Neglecting some corrections, it is given by:
Qs(Y ) = Q0exp(αsλY )Y −β λ ≈ 2.4
• The saturation scale is the dividing line between dense and dilute regions.
• The higher the rapidity the denser the system gets and partons start to re-interact.
• Also, saturation occurs earlier if the size of partons is bigger.
BK, E.G. de Oliveira–GFPAE – p. 29
Saturation scale
• Dilute and saturated regimes can be represented in 2-D graphs (Y = ln(1/x)):
Q1r= −
Y
sQ (Y)
Q1r= −
Y
Non−perturbative
Saturation scale
Dense system
region
Dilute systemLinear evolution
Nonlinear evolution
BK, E.G. de Oliveira–GFPAE – p. 30
GBW model
• Linear growing and saturation properties of BK equation are roughly similarlyfound in the Golec-Biernat and Wusthoff (GBW) saturation model.
• One of the key ideas to GBW model is the assumption of a x-dependent radius:
R0(x) =1
Q0
„
x
x0
«λ
2
, (18)
which scales the quark-antiquark separation r in the dipole cross section:
r =r
2R0(x). (19)
• Then, the dipole cross section can be written as:
σdip(x, r2) = σ0g(r2). (20)
• Q0 = 1 GeV sets the dimension.
BK, E.G. de Oliveira–GFPAE – p. 31
GBW model: dipole cross section
• The function g(r2) chosen is:
g(r2) = 1 − exp[−r2]. (21)
• The dipole cross section is then:
σdip(x, r) = σ0
1 − exp
"
−„
r
2R0
«2#!
= σ0
„
1 − exp
»
−r2Q20
4
“x0
x
”λ–«
(22)
• The three parameters were fitted to σ0 = 23mb, λ = 0.288 and x0 = 310−4.
• The cross section can be written as a function of Y also:
σdip(Y, r) =
Z
d2bN(b, r, Y ) = σ0
»
1 − exp
„
− r2Q2
s(Y )
4
«–
with Q2s(Y ) = exp(0.28(Y − Y0)).
BK, E.G. de Oliveira–GFPAE – p. 32
GBW model: σdip behavior
The cross section behavior is shown in the figure:
0 0.5 1 1.5 2r (GeV)
0
10
20
30
σ dip
(x,r
) (m
b)GBW
BK, E.G. de Oliveira–GFPAE – p. 33
Saturation and color transparency
• In GBW model, when r << 1/Q2s(Y ), the phenomenon of color transparency
appears (small dipoles have small chance of interaction):
σ(Y, r)
σ0=r2Q2
s(Y )
4.
• When r >> 1/Q2s(Y ), the cross section saturates:
σ(Y, r)
σ0≈ 1.
BK, E.G. de Oliveira–GFPAE – p. 34
GBW Results
The γ∗p-cross section for various energies (solid lines: quark mass of 140MeV; dottedlines: zero quark mass).
W=245 (x128)
W=210 (x64)
W=170 (x32)
W=140 (x16)
W=115 (x8)
W=95 (x4)
W=75 (x2)
W=60 (x1)
ZEUS BPC
H1
H1 low Q 2
Q2 (GeV2)
σ γ*p (
µb)
Q2 (GeV2)
σ γ*p (
µb)
1
10
10 2
10 3
10 4
10-2
10-1
1 10 102
BK, E.G. de Oliveira–GFPAE – p. 35
GBW Results
The results for the fit to the inclusive HERA data on F2 for different values of thevirtuality.
H1/ZEUS
F2
Q2=1.5 GeV2 Q2=2 Q2=2.5 Q2=3.5 Q2=4.5 Q2=5
Q2=6.5 GeV2 Q2=8.5 Q2=10 Q2=12 Q2=15 Q2=18
Q2=20 GeV2 Q2=22 Q2=25 Q2=27 Q2=35 Q2=45
Q2=60 GeV2 Q2=70 Q2=90 Q2=120 Q2=150 Q2=200
Q2=250 GeV2 Q2=350 Q2=450 Q2=650 Q2=800
H1(94/95)
ZEUS(94)
x
0
0.5
1
1.5
0
0.5
1
1.5
0
0.5
1
1.5
0
0.5
1
1.5
0
0.5
1
1.5
10-4
10-3
10-2
10-4
10-3
10-2
10-4
10-3
10-2
10-4
10-3
10-2
10-4
10-3
10-2
10-4
10-3
10-2
BK, E.G. de Oliveira–GFPAE – p. 36
Saturation without unitarization
• Consider impact parameter dependence.
• Total inelastic cross section, in the saturation regime (t = ln(s/s0)):
σ = πR2(t) . (23)
• To satisfy the Froissart bound the radius R(t) should grow at most linearly with t.
• However, at large rapidities the radius of the saturated region is exponentially large
R(t) = R(t0) exphαsNc
2πǫ(t− t0)
i
. (24)
[A. Kovner, U. A. Wiedemann; Phys.Rev. D66 (2002) 051502.]
BK, E.G. de Oliveira–GFPAE – p. 37
Saturation without unitarization
BK, E.G. de Oliveira–GFPAE – p. 38
Geometric scaling
• Geometric scaling is a phenomenological feature of DIS which has been observedin the HERA data on inclusive γ∗ − p scattering, which is expressed as a scalingproperty of the virtual photon-proton cross section
10-1
1
10
10 2
10 3
10-3
10-2
10-1
1 10 102
103
E665
ZEUS+H1 high Q2 94-95H1 low Q2 95ZEUS BPC 95ZEUS BPT 97
x<0.01
all Q2
τ
σ totγ*
p [µ
b]
σγ∗p(Y,Q) = σγ∗p (τ) , τ =Q2
Q2s(Y )
where Q is the virtuality of the photon,Y = log 1/x is the total rapidity andQs(Y ) is the saturation scale
[Stasto, Golec Biernat and Kwiecinsky,2001]
BK, E.G. de Oliveira–GFPAE – p. 39
Geometric scaling
• Looking to the solutions of BK equation, one sees that they reach a universalshape independently of the initial condition.
• The solutions even look similar when calculated for different Y (they appear to beshifted when Y is changed).
• In terms of the scattering amplitude, the geometric scaling means that BK solutiondepends only on rQs(Y ) for asymptotic values of rapidity.
N (r, Y ) = N (rQs(Y )).
• Using Qs(Y ) ≈ Q0exp(αsλY );
N (r, Y ) = N (Q0exp(lnr + αsλY ))
and the geometric scaling relation resembles as a traveling wave with velocity αsλ,time Y and spatial coordinate lnr.
BK, E.G. de Oliveira–GFPAE – p. 40
Geometric scaling
√τσγ∗p plotted versus
the scaling variable τ .
10-1
1
10
10 2
10-3
10-2
10-1
1 10 102
103
E665
ZEUS+H1 high Q2 94-95
H1 low Q2 95ZEUS BPC 95ZEUS BPT 97
x<0.01
all Q2
τ
√τ– σto
tγ*p
[µb]
BK, E.G. de Oliveira–GFPAE – p. 41
Geometric scaling
Different values of scal-ing variable versus ex-perimental data.
10-2
10-1
1
10
10 2
10-6
10-5
10-4
10-3
10-2
0.0075
0.0178
0.042
0.1
0.24
0.56
1.33
3.15
7.47 17.7 42.0 100
HERA fixed target
x
Q2 (G
eV2 )
BK, E.G. de Oliveira–GFPAE – p. 42
Geometric scaling
Experimental data onσγ∗p from the regionx > 0.01 plotted ver-sus the scaling variableτ = Q2R2
0(x).
10-3
10-2
10-1
1
10
10 2
10-1
1 10 102
103
104
105
H1+ZEUS
BCDMSNMCE665EMCSLAC
x>0.01
all Q2
τ
σ totγ*
p [µ
b]
BK, E.G. de Oliveira–GFPAE – p. 43
BK equation in momentum space
• Performing the Fourier transform:
N (k, Y ) =
Z
d2r
2πe−i~k·~r N (r, Y )
r2, (25)
BK equation is given by:
dN (k, Y )
dY= αs
Z
dk′
k′K(k, k′)N (k′, Y ) − αsN 2(k, Y ). (26)
• The solution to the linear part (BFKL) in saddle point approximation is:
kN (k, Y ) =1
p
παsχ′′(0)Yeαsχ(0)Y exp
− ln2(k2/k20)
2αsχ′′(0)Y
!
(27)
with χ(0) = 4 ln 2 and χ′′(0) = 28ζ(3).
• The last term in the expression above presents the diffusion.
BK, E.G. de Oliveira–GFPAE – p. 44
BK equation in momentum space
0.5
1
1.5
2
2.5
3
10-5
10-4
10-3
10-2
10-1
1 10 102
103
104
BFKL
Kovchegov
k [GeV]
k φ(
k,y)
BK, E.G. de Oliveira–GFPAE – p. 45
BK equation in momentum space
• BFKL presents strong diffusion and can be interpreted as a random walk in ln k
with rapidity as time.
• The Gaussian shapes of last figure are then expected, as well as the widthincrease of distributions.
• The initial condition used was
kN (k, Y = 0) = δ(k) (28)
• One sees that BK equation leads to a suppression of the diffusion into infrared.
• Therefore, one defines the saturation scale as the distribution peak positionQs(Y ) = kmax(Y ).
• Other way to see the same effects is using the distribution
Ψ(k, Y ) =kN (k,Y )
kmax(Y )N (kmax(Y ),Y ):
BK, E.G. de Oliveira–GFPAE – p. 46
BK equation in momentum space
0
1
2
3
4
5
6
7
8
9
10
-5 -4 -3 -2 -1 0 1 2 3 4 5
αs = 0.2
Balitsky-Kovchegov
BFKL
log10(k/1GeV)
log 10
(1/x
)
BK, E.G. de Oliveira–GFPAE – p. 47
BK equation in momentum space
• In the last figure, we see that BK equation shifts the contour plot towards highervalues of transverse momenta.
• Two regions are found: in the first, there is still diffusion, particularly for large k.
• In the second, diffusion is suppressed and the contour lines are parallel.
• This is an indication of geometric scaling, since straight lines can be parametrizedby ξ = ln k/k0 − λY + ξ0.
• One sees also that for large k BK and BFKL equations present similar solutions,but BFKL presents an unlimited increase with energy. On the other side, BKsolutions are bounded.
BK, E.G. de Oliveira–GFPAE – p. 48
Traveling waves
• The BK equation can be compactly written in momentum space also as
∂Y N (ρ, Y ) = αχ(−∂ρ)N − αN 2. (29)
• The Mellin–transformed BFKL kernel χ(γ) is given by:
χ(γ) = 2ψ(1) − ψ(γ) − ψ(1 − γ), with ψ(γ) =Γ′(γ)Γ(γ)
. (30)
• χ(−∂ρ) is an integro-differential operator defined with the help of the formal seriesexpansion:
χ(−∂ρ) = χ(γ0)1 + χ′(γ0)(−∂ρ − γ01) +1
2χ′′(γ0)(−∂ρ − γ01)2
+1
6χ(3)(γ0)(−∂ρ − γ01)3 + . . . (31)
for some γ0 between 0 and 1, where we used the identity operator 1.
BK, E.G. de Oliveira–GFPAE – p. 49
BK and FKPP equations
• Munier and Peschanski showed that after the change of variables
t ∼ αY, x ∼ log(k2/k20), u ∼ φ, (32)
• The approximation of BFKL kernel by the first three terms of the expansion:
χ(−∂ρ) = χ(γ0)1 + χ′(γ0)(−∂ρ − γ01) +1
2χ′′(γ0)(−∂ρ − γ01)2,
• And a saddle point approximation:
χ(−∂ρ) = −χ′(γc)∂ρ +1
2χ′′(γc)(−∂L − γc1)2,
• BK equation is reduced to the Fisher and Kolmogorov-Petrovsky-Piscounov(FKPP) equation from non-equilibrium statistical physics:
∂tu(x, t) = ∂2xu(x, t) + u(x, t) − u2(x, t), (33)
where t is time and x is the coordinate. FKPP dynamics is called reaction-diffusiondynamics.
BK, E.G. de Oliveira–GFPAE – p. 50
Traveling waves
• FKPP equation admits so-called traveling wave solutions.
• The position of a wave front x(t) = v(t)t for a traveling wave solution does notdepend on details of nonlinear effects.
• At large times, the shape of a traveling wave is preserved during its propagation,and the solution becomes only a function of the scaling variable x− vt.
x
u(x, t)
0
1
t = 0 t1
X(t1)
2t1
X(t2)
3t1
X(t3)
BK, E.G. de Oliveira–GFPAE – p. 51
Traveling waves and saturation
Y = 25Y = 20Y = 15Y = 10Y = 5Y = 0
log(k2)
T
4035302520151050-5
10
1
0.1
0.01
0.001
1e-04
1e-05
• In the language of saturation physics the position of the wave front is nothing butthe saturation scale x(t) ∼ lnQ2
s(Y ) and the scaling corresponds to the geometricscaling x− x(t) ∼ ln k2/Q2
s(Y ).
• Summarizing: time t→ Y ; space x→ L; wave front u(x− vt) → N (L− vY ) andtraveling waves → geometric scaling.
BK, E.G. de Oliveira–GFPAE – p. 52
Traveling waves and saturation
• In the dilute regime, in which one has
T (k, Y )k≫Qs≈
„
k2
Q2s(Y )
«−γc
log
„
k2
Q2s(Y )
«
exp
"
− log2`
k2/Q2s(Y )
´
2αχ′′(γc)Y
#
, (34)
where
λ = min αχ(γ)
γ= α
χ(γc)
γc= αχ′(γc), α ≡ αsNc
π. (35)
• Geometric scaling is obtained within the window
log`
k2/Q2s(Y )
´
.p
2χ′′(γc)αY . (36)
BK, E.G. de Oliveira–GFPAE – p. 53
Next-to-leading order BK
• As usual, to get the region of application of the leading-order evolution equationone needs to find the next-to-leading order (NLO) corrections.
• Unlike the DGLAP evolution, the argument of the coupling constant in LO BKequation is left undetermined in the LLA.
• Careful analysis of this argument is very important from both theoretical andexperimental points of view.
• Balitsky calculated the quark contribution to NLO B equation Phys.Rev. D75,014001 (2007).
• Balitsky and Chirilli calculated the gluon contribution to NLO B equation Phys.Rev.D77, 014019 (2008).
BK, E.G. de Oliveira–GFPAE – p. 54
Next-to-leading order BK
• NLO result does not lead automatically to the argument of coupling constant infront of the leading term.
• In order to get this argument, it was used the renormalon-based approach.
• The running coupling was roughly found to be αs(|x− y|), but:
αs((x−y)2)
2π2
(x−y)2
X2Y 2 |x− y| ≪ |x− z|, |y − z|αs(X2)
2π2X2 |x− z| ≪ |x− y|, |y − z|αs(Y 2)
2π2Y 2 |y − z| ≪ |x− y|, |x− z| (37)
BK, E.G. de Oliveira–GFPAE – p. 55
Next-to-leading order BK
• Next-to-leading order BK:
• X = x− z, Y = y − z, X = x− z′, Y = y − z′.
• As we can infer, in the complete hierarchy, the single dipole evolution is related tothe sextupole terms.
BK, E.G. de Oliveira–GFPAE – p. 56
Beyond BK equation: Pomeron loops
BK, E.G. de Oliveira–GFPAE – p. 57
Beyond BK equation: Pomeron loops
BK, E.G. de Oliveira–GFPAE – p. 58
Beyond BK equation: Pomeron loops
BK, E.G. de Oliveira–GFPAE – p. 59
Beyond BK equation: Fluctuations
• BK equation completely misses the effects of fluctuations in the gluon (dipole)number, related to discreteness of the evolution.
• Also, since BK equation uses a mean field approach, < T 2 > − < T >2= 0.
• After an approximation related to the impact parameter dependence, it can beshown that a Langevin equation for the event-by-event amplitude can rebuild thePomeron loop hierarchy.
• This is formally the BK equation with a Gaussian white noise term.
• It lies in the same universality class as the stochastic FKPP equation (sFKPP):
∂tu(x, t) = ∂2xu(x, t) + u(x, t) − u2(x, t) +
r
2
Nu(x, t)(u(x, t) − 1)ν(x, t). (38)
• White noise is defined as:
< ν(x, t) >= 0 < ν(x, t)ν(x′, t′) >= δ(t− t′)δ(x− x′). (39)
[E. Iancu, D.N. Triantafyllopoulos; Nucl.Phys.A 756, 419 (2005)]
BK, E.G. de Oliveira–GFPAE – p. 60
Fluctuations
• Each realization of the noise means a single realization of the target in theevolution, which is stochastic, and this leads to an amplitude for a single event.
• Different realizations lead to a dispersion in the solutions and also in the saturationmomentum ρs ≡ ln(Q2
s(Y )/k20).
• For each single event, the evolved amplitude shows a traveling-wave pattern, withspeed of the wave smaller than the value predicted by BK equation:
λ∗ ≃ λ− π2γcχ′′(γc)
2 ln(1/α2s)
(40)
• The saturation scale Qs(Y ) is now a random variable whose average value isgiven by
〈Q2s(Y )〉 = exp [λ∗Y ]. (41)
• The dispersion in the position of the individual fronts is given by
σ2 = 〈ρ2s〉 − 〈ρs〉2 = DαY, (42)
BK, E.G. de Oliveira–GFPAE – p. 61
Diffusive scaling
• Geometric scaling:
〈T (ρ, ρs)〉 = T (ρ− 〈ρs〉) . (43)
is replaced by diffusive scaling
〈T (ρ, ρs)〉 = T„
ρ− 〈ρs〉√αDY
«
. (44)
BK, E.G. de Oliveira–GFPAE – p. 62
Fluctuations
BK, E.G. de Oliveira–GFPAE – p. 63
Toy model• Are fluctuations really important (in the context of a toy model)?• At fixed coupling they are [E. Iancu, J.T. de Santana Amaral, G. Soyez, D.N.
Triantafyllopoulos; Nucl.Phys. A786, 131 (2007)].• However, at running coupling they are not [A. Dumitru, E. Iancu, L. Portugal, G.
Soyez, D.N. Triantafyllopoulos; JHEP 0708:062 (2007)].• This reflects the slowing down of the evolution by running coupling effects, in
particular, the large rapidity evolution which is required for the formation of thesaturation front via diffusion.
0
0.5
1
1.5
2
2.5
3
3.5
4
-10 0 10 20 30 40
T(x
-xs,
Y)
eγ s(x
-xs)
x-xs
MFA fluct.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
-10 -5 0 5 10 15 20 25
T(x
-xs,
Y)
eγ s(x
-xs)
x-xs
MFA fluct.
BK, E.G. de Oliveira–GFPAE – p. 64
References
Y.V. Kovchegov, Phys. Rev. D 60 (1999) 034008; Phys. Rev. D 61 (2000) 074018.I. Balitsky, Nucl. Phys. B 463 (1996) 99.A.M. Stasto, Acta Phys. Polonica 35 (2004) 3069.C. Marquet, G. Soyez, Nucl. Phys. A 760 (2005) 208.K. Golec-Biernat, M. Wüsthoff, Phys. Rev. D 59 (1999) 014017; Phys. Rev. D 60 (1999)114023; Eur. Phys. J. C 20 (2001) 313.S. Munier R.Peschanki, Phys. Rev. Lett. 91 (2003) 232001; Phys. Rev. D 69 (2004)034008; hep-ph/0401215.R.A. Fisher, Ann. Eugenics 7 (1937) 355. A. Kolnogorov, I. Petrovsky, N. Picounov,Moscow Univ. Bull. Math. A 1 (1937) 1.M.B. Gay Ducati, V.P. Goncalves, Nucl. Phys. B 557 296.A.M. Stasto, K. Golec-Biernat and J. Kwiecinsky, Phys. Rev. Lett. 86 (2001) 596;hep-ph/0007192.
BK, E.G. de Oliveira–GFPAE – p. 65
AGL equation from BK equation
• AGL equation can be derived from BK equation. Taking the equation derived byKovchegov:
N (x01,b0, Y ) = −γ(x01,b0) exp
»
−4αsCF
πln
„
x01
ρ
«
Y
–
+αsCF
π
Z Y
0dy exp
»
−4αsCF
πln
„
x01
ρ
«
(y − Y )
–
×Z
ρd2x2
x201
x202x
212
[2N (x02,b0, Y −N (x02,b0, Y )N (x12,b0, Y )].(45)
• In the double logarithmic limit, in which the momentum scale of the photon Q2 islarger than the momentum scale of the nucleus ΛQCD, the large Q2 limit of lastequation reduces to:
∂N (x01,b0, Y )
∂Y=
αsCF
πx201
Z 1/Λ2QCD
x201
d2x02
x202x
212
[2N (x02,b0, Y
−N (x02,b0, Y )N (x02,b0, Y )]. (46)
BK, E.G. de Oliveira–GFPAE – p. 66
AGL equation from BK equation
• The last equation can be derived again with respect to ln(1/x201Λ
2QCD):
∂N (x01,b0, Y )
∂Y ∂ ln(1/x201Λ2
QCD)=
αsCF
π[2 −N (x01,b0, Y )]N (x01,b0, Y ). (47)
• In the physical picture for the dipole evolution in the DLA limit, the produceddipoles at each step of the evolution have much greater transverse dimensionsthan the parent dipoles.
• The connection in this approximation between N and the nuclear gluon function is:
N (x01,b0 = 0, Y ) = 2
1 − exp
»
−αsCF π2
N2c S⊥
x201AxG(x, 1/x2
01)
–ff
. (48)
• From the above equations and using x01 ≈ 2/Q, one obtains:
∂N (x01,b0, Y )
∂Y=
αsCF
πx201
Z 1/Λ2QCD
x201
d2x02
x202x
212
[2N (x02,b0, Y
−N (x02,b0, Y )N (x02,b0, Y )]. (49)
• This equation is precisely the AGL equation.BK, E.G. de Oliveira–GFPAE – p. 67
Light-cone kinematics
• Let z be the longitudinal axis of the collision. For an arbitrary 4-vectorvµ = (v0, v1, v2, v3), the light-cone (LC) coordinates are defined as
v+ ≡ 1√2(v0 + v3), v− ≡ 1√
2(v0 − v3), v ≡ (v1, v2) (50)
• One usually writes v⊥ ≡ |v| =p
(v1)2 + (v2)2
• In these coordinates, x+ ≡ 1√2(t+ z) is the LC time and x− ≡ 1√
2(t− z) is the
LC longitudinal coordinate.
• The invariant scalar product of two 4-vectors reads
p · x = p0x0 − p1x1 − p2x2 − p3x3
=1
2(p+ + p−)(x+ + x−) − 1
2(p+ − p−)(x+ − x−) − p · x
= p−x+ + p+x− − p · x (51)
• This form of the scalar product suggests that p− should be interpreted as the LCenergy and p+ as the LC longitudinal momentum.
BK, E.G. de Oliveira–GFPAE – p. 68
Light-cone kinematics
• For particles on the mass-shell, k± = (E ± kz)/√
2, with E2 = (m2 + k2z + k2)
k+k− =1
2(E2 − k2
z) =1
2(k2 +m2) ≡ m2
⊥ (52)
• One needs also the rapidity
y ≡ 1
2lnk+
k−=
1
2ln
2k+2
m2⊥
(53)
• Under a longitudinal Lorentz boost (k+ → βk+, k− → (1/β)k−, with constant β),the rapidity is shifted only by a constant, y → y + β
• For a parton inside a right-moving (in the positive z direction) hadron, we introducethe boost-invariant longitudinal momentum fraction x
x ≡ k+
P+(54)
BK, E.G. de Oliveira–GFPAE – p. 69